If the host name is changed by a manifest run, the old host name is used as node names

[digimer]

Which then breaks scan-cluster because it can't map the node name to a host name/uuid. Renaming the nodes requires stopping the cluster and updating corosync.conf, hosts and pcs hosts.

# Rename pacemaker node name
[root@an-a02n01 ~]# pcs cluster stop --all
vm-a02n02: Stopping Cluster (pacemaker)...
vm-a02n01: Stopping Cluster (pacemaker)...
vm-a02n02: Stopping Cluster (corosync)...
vm-a02n01: Stopping Cluster (corosync)...
[root@an-a02n01 ~]# vim /etc/corosync/corosync.conf
[root@an-a02n01 ~]# vim /etc/hosts 
[root@an-a02n01 ~]# pcs host auth an-a01n01 an-a01n02 -u hacluster
Password: 
an-a01n02: Authorized
an-a01n01: Authorized
[root@an-a02n01 ~]# 

[digimer]

This also breaks stonith, requiring existing stonith methods be deleted (so scan-cluster and recreate under the new name).